Vancouver Sun
A curious two days at the . Nine speakers panels over the two days – in essence tracing sockeye from the time they rise out of the gravel, through their various life stages until they return back to the Fraser. The “Summit” was also a follow-up to the “Salmon Think Tank” hosted in early Dec. 2009. From that think tank came a press statement.
. I have commented on this press statement in some of the first on this site.
One of the points I made on earlier posts, which I still maintain is something that we can demonstrate in another “graph”:
This is from Jessica Hagy’s blog .
I’m not sure, though, that when it comes to salmon when we might have reached the bottom of the curve where confusion was the least — maybe prior to European contact…?
As I mentioned in yesterday’s post, this Summit was largely a scientific exercise — and fair enough, it was hosted by Simon Fraser University Centre for Coastal Studies. The folks there did a nice job of pulling this together and have continued to do a nice job over many years hosting the Speaking for the Salmon series of events.
However, it was also not all scientists in the over 100 folks attending, and yesterday — day two — shifted a little more into some curious discussions such as “putting a value on salmon”, “what actions can be taken?”, “what tools do we have?”, and “how do we move forward?”. In these sessions, the voices of some commercial fishermen and community activists started to surface more — and some strong messages that parallel my post from yesterday: managing fish, especially salmon, is not really a scientific exercise; it’s a political exercise.
This is evident around the world in the decline of fisheries. One does not have to look much further than the iconic Bluefin Tuna and the challenges of protecting this species in the face of serious declines.
As I lay in bed trying to fall asleep last night I pondered the many perspectives that surfaced ranging along a spectrum of opinion about what happens next. That’s probably a book in itself… however, one of the strong notions that continues to surface in my mind is that “fisheries science” also ranges along a spectrum of opinion. Sure it’s opinion backed by a lot of charts and graphs, and various letters behind people’s name, and empirical methods and so on — however, as was made clear on the first day of this summit there are as many “models” for attempting to predict fish populations and patterns (especially salmon) as there are models in a Sears Christmas Wishbook or Victoria Secret catalogue.
Now, I mean no offence to learn-ed individuals that have spent years in hallowed institutions counting, measuring, tagging, and chasing fish. Many of these individuals make valuable contributions to the discussion.
Yet, in my mind, after looking at so many differing charts and graphs and hearing that salmon scientists have their own meetings where they have “salmon pools” — like a hockey pool. One presentation explained how various “salmon scientists” met this past year and a bottle of wine was wagered for the person who had the most accurate salmon forecasting model.
The most entertaining aspect of this idea is how the hell does anyone confirm who was right; who wins the bottle of wine?
To accurately “forecast” salmon returns, one would need an accurate count of how many salmon actually returned to confirm the forecast.You know, like weather… when we forecast rain, and it doesn’t rain; forecast was wrong. Not that this ever happens…
To count how many salmon actually return is a highly variable exercise — especially in the Fraser River. Counting salmon that actually reach spawning grounds is barely even guess work at best. Some rivers are flown by helicopter or otherwise over a period of a few days, some rivers use mark-recapture methods to extrapolate (i.e. estimate) over the whole run, other streams are walked every few days, and many streams are not even looked at because of sheer numbers of streams and large geographic area.
I’m thinking maybe that bottle of wine should be put in someone’s cellar and opened in a hundred years. The bottom line is that when it comes to wild salmon — the “science” when it comes to numbers, is largely guess-work; estimates; opinions.
It is one part endless charts and graphs, one part chasing rainbows, and another part endless computer modeling. Throw in a dash of guess-timate, a dollop of estimate, and a whole lot of mystery.
After seeing how many different “models” exist out there to try and estimate salmon returns and estimate salmon populations from the time they leave the gravel, migrate through the North Pacific, and return to their home streams I was reminded of the story I’ve heard about how salmon arrived on Haida Gwaii (sometimes referred to as the Queen Charlotte Is).
Haida Raven Decal - Billy Bedard
As I’ve heard the story, Raven brought salmon to Haida Gwaii. On a visit to the mainland, Raven in his trickster ways managed to roll up rivers, lakes and streams and the salmon they contained — into his beak — fly back over the Hecate Strait and drop them onto the islands. (the story is so much more interesting then my short paraphrase)
Throughout the historic range of wild salmon are aboriginal stories of how salmon came to be in those lands. Often these stories aren’t too far separated from the stories of human creation — they often involve Raven or Coyote or other supernatural creator beings.
I find more solace in those stories than I do in the “scientific” method. It seems that those stories guided humans for thousands of years in how they co-existed and co-evolved with wild salmon. It seems that those stories guided a sustainable relationship — sure there were some hard times; however, those stories are generally few and far between.
Yet, in a mere hundred years or so, “scientific” methods of “fisheries management” have taken us down a road of fisheries declines and collapses the world over. And, not just science — but worse yet, the political decisions on top of the science.
Curiously, I looked up the definition of science on Wikipedia and this is what it states:
Science (from the Latin scientia, meaning “knowledge”) is, in its broadest sense, any systematic knowledge-base or prescriptive practice that is capable of resulting in a correct prediction, or reliably-predictable type of outcome. In this sense, science may refer to a highly skilled technique, technology, or practice, from which a good deal of randomness in outcome has been removed.
When it comes to science and wild salmon… we have not, and most likely will never, remove randomness in outcome. Yes, there are systematic knowledge based practices, even some prescriptive practices — however often resulting in poor predictions, and far from reliably predicted outcomes.
Once we can get past the thought that “science” is going to solve this one for us; once we stop saying “let’s delay action until we do more studies”; once we get past wavering politicians with the inability to make brave decisions on wild salmon; once we get past looking for the “smoking gun”; once we get past the largely useless exercise of expensive public inquiries that keep saying the same thing — then maybe we can take more fricking action.
Taking action is an individual choice — not a scientific one.
Hello Salmonguy,
Mark-recapture does not extrapolate. Extrapolate means to estimate outside of the tabulated range or observed range. It is an estimate based on the fulfillment of 5 assumptions (tagged and untagged fish have a equal opportunity of capture and recapture; untagged fish behave the same as tagged fish; a closed system, tags are not missed; fish do not lose their tags) being satisfied. Whether these assumptions are statisfied or violated is part of post-season data analysis (bias testing) which is highly scrutinized. These analyses are carried out by dedicated people who also understand fish (many are anglers themselves) – not just the stats behind the analysis. When biases are found, feedback is given pre-season to improve the study design to remove or, in most cases, reduce the biases. Mark-recaptures have also been recently calibrated with other high precision methods (DIDSON sonar or Dual Frequency Identification Sonar) and have been within 95% confidence limits.
It is important to note that I mention the word “precision” and not accuracy. It would be most beneficial to get an absolute count which is done at some fence projects in the Fraser River watershed (i.e. Stellako, Eagle, Birkenhead and Cultus), but this is not realistic most of the time. Precision is deemed the most important and where most emphasis is place. I will use the dart board analogy. You can be accurate one time on the board, but if your darts are all over the dart board the next then that is no good at all. A tight grouping as close to the bullseye is more important, not to mention more attainable and realistic then hitting the bullseye every shot.
Visual surveys such a ground counts on foot or helicopter are examples of low precision methods while mark-recapture, DIDSON, fence are examples of high precision methods. There is a little more to it than what you have suggested. As you can appreciate, it is not possible to obtain an absolute count of these fish from such a huge watershed. Dollars, environmental conditions, time or year, the biology of the fish and the need for a low or high precision method often dictates what method of enumeration is applied. It would be nice to have high precision on all projects, but when you understand what is involved with a particular method of enumeration you will soon see its limitations and advantages.
Methodologies have changed over the years to make things better. Post season review and crew feedback have been used to improve methodologies, but more has to be done I admit. Anyone that does this type of work (whether it is in the field or in the office) is always looking to improve the way things are done. There are people that do take some pride in their work. For instance, recent work has been done to improve expansion indices in the watershed. It is important to note that future improvements in enumeration methodologies should not be seen as just a failure of the past, but rather a constant refinement that is always ongoing. Technology is much different than it was 30 years ago. DIDSON is now being used – which is different from traditional split beam technology. Training in different technologies have also changed.
I realize that people are frustrated with science-end of things, but part of the problem is the communication of results rather than the result itself. From my perspective, the general public likes certainty and an absolute valve, but this is not the reality of science. People (Not you in particular…I am speaking generally) that indicate that this is easily obtainable or critisize those that undertake this work are often misinformed.
If you get the chance you should visit an enumeration project this fall (i.e. Adams) and see how a project like this conducted. A project like a mark-recapture is not a small undertaking and much effort is involved.
Brian, thanks for the thoughtful comment and time spent to formulate it.
I appreciate where you are coming from and have a lot of respect for folks out there with their hands and feet in the streams working on these projects. And yes, I’ve certainly simplified things to a certain degree; however, I’ve also worked on various ‘counting’ projects — from visual counts at fences, to mark-recapture, to snorkel surveys, to various catch-per-unit-of-effort (CPUE), and so on.
I’ve also chatted with various folks about tools such as DIDSON and sonar and have heard a range of opinions — from the 95% confidence limits you suggest to “voodoo science at best”. Some common issues identified are double counting fish that swim upstream then back downstream, not being able to discern between other species (salmon and other fish), and observer error, and limits to counting across an entire channel, or just on one side, and so on. Also, with the DIDSON projects, on the Horsefly River for example, data capacity constraints limited to only counting for 20 minutes and then ‘extrapolating’. (all challenges I’m sure you’re well aware of).
I’m not too sure I agree with your definition of “extrapolate” – by definition it means: “To infer or estimate by extending or projecting known information” or in Mathematics to “estimate (a value of a variable outside a known range) from values within a known range by assuming that the estimated value follows logically from the known values.” I could be wrong, however, I see the various methods of counting fish as pretty much that: i.e. “here’s what we know (i.e. counted, or re-caught, etc.) and now let’s extrapolate that over time, or over a geographic area”.
I often get a chuckle out of the idea of “95% confidence”, or accuracy, or other ways of framing the fact that we don’t really know – and we’ll never really know (unless we are dealing with very small numbers on very small streams). How can we be 95% confident that a count is correct? And how big is that 5% of non-confidence?
If a race car driver does his/her job with 95% confidence, or 95% accuracy (I recognize these are different things), that 5% of non-confidence or 5% of error can result in death. (and this is not just race car drivers… anyone driving at any time).
As you mention there are constraints to counting every fish — even fence counts are notoriously off by some degree of error — and thus we can never actually know precisely what our accuracy levels are because we are unable to say: “well our equations suggest there are this many fish, and we know from counting every last fish that there were actually this many fish in the creek, and thus our error is n.”
Foot counts, even on small streams are highly inaccurate – esp. when one is trying to count the elusive, darkly colored coho that loves to hide under anything possible – or just zip in at night spawn and then gone (or eaten). Plus some streams are too deep, too turbid, turbid one day and not the next, full of logs or other debris, and so on…
Worse yet, since our counting of actual spawners is reflected along a range of confidence and actual accuracy — and then we (or fish “managers”) base harvest decisions on these varying ranges of counts (i.e. Maximum Sustainable Yield – which resulted in 80% of Fraser sockeye being taken for over 50 years) — we then have a problem (as we are seeing these days).
The choice of “precision” over “accuracy” is a very curious one… precision has two different meanings (and I don’t say this for the sake of argument, just that I have some fascination with word definitions and how we choose to use those words):
1. The state or quality of being precise; exactness. (and this we know we are not when it comes to counting salmon; its impossible to be exact).
2. The ability of a measurement to be consistently reproduced. (and this is probably closer to the meaning that is used in counting fish).
I think definition 2 is probably closer to fisheries management and science. Sure the measurement can be consistently reproduced (i.e. mark-recapture with its many assumptions); however, how accurate or precise is it? We really don’t know, because we don’t have a 100% accurate actual count to compare it to. In essence, everything is still an estimate — not a precise count like say: I have four fingers and two thumbs…
And so, I appreciate your observation about frustration with the “science-end” of things. I completely agree that communication of results is one part of the problem; however I don’t necessarily agree with the idea that the public is looking for “certainty” or an “absolute valve” — and I think this gets to the heart of some of the issues. I think, what more of the general public would appreciate is folks on the science-end-of-things to speak in more absolutes. Meaning… instead of scientists trying to suggest that we “do know” (a favorite past time in DFO); clearly express that we don’t really know, and that we try and do the best that we can with what we’ve got.
And if this is the case, a much clearer articulation of the “precautionary principle” — not just in definition, but in action.
I’ve started to hear more of this on the forecasting side of things — especially after last years debacle on Fraser sockeye. Now, several scientists are pushing to make it clear that forecasting salmon runs is a highly, highly imprecise process. The comparison is drawn to weather forecasting (e.g. Randall Peterman at SFU). With weather, meteorologists are relatively accurate over 2 days; less so over 7 days; and largely in a crap-shoot over 14 days. And this is with billions of dollars of tools and expertise at hand.
Salmon forecasting… not so much.
Thus, if:
1. more fisheries scientists would explain the imprecise nature of the science and that we are simply trying our best with what we’ve got — rather than many scientists living on their little islands of expertise communicating results in charts, graphs, and equations that the general public has no idea what’s going on. And, simply publishing results in “peer-reviewed” journals… that means exactly that… largely only one’s peers are reading it.
2. Take a stand on something. I heard a senior U.S. bureaucrat express this in no uncertain terms last week in Portland. I paraphrase: “take a bloody stand on something… I don’t want to hear: oh, maybe this, or maybe that, or maybe that other thing… tell me what you know, tell me what you don’t know and then we can make some decisions…”.
I can wholeheartedly agree with this type of statement. There’s nothing more frustrating then, for example, reading all of the studies coming out of Carnation Creek on Vancouver Island in the 80s and 90s regarding clearcutting in salmon streams and having folks suggest there was probably no impact. Of course there’s an impact… what’s happening is an economic trade-off between the logging industry and healthy salmon streams… (and some other factors).
Of course, scientists taking a stand on things begins to threaten the great monolith of “objectivity” — this has certainly been thrown towards Alexandra Morton and her work as a scientist, and as an activist. These, in many scientists eyes, are mutually exclusive endeavors…
I often point to the great paradox of the properties of light. Depending on the experiment, light either displays particle properties, or it displays wave properties — but never both at the same time. So no matter how much an expert on one particular aspect of light, a scientist on one side of the debate proves their point in repeatable, measurable results, they are always ‘wrong’ — because they are missing half the equation, half the reality… With salmon – and how fisheries science approaches them — we are missing far more than half the equation. We call it fisheries science because it focuses on the fish; bringing in the rest of the ecosystem (e.g. North Pacific) is impossible.
And so, as with everything: light is both particle and wave… as with the salmon debates, they need to be both science and activism and politics and economics, and so a huge component of “we have no frigging idea…”
I certainly don’t raise the points to argue, or fully disagree, more to add to an important discussion that seems to be occurring a lot more frequently, and over a wider area. In the end if anything is to improve for salmon, everyone needs to be involved: public, policy makers, politicians, scientists, and so on.
thanks again for adding to the comments. much appreciated.
Hello Salmonguy,
Thank you for your comments. It is nice to hear that you have had some exposure to fisheries work; however, I am responding back (just read your reply now actually) to clear up a few things regarding DIDSON.
First, I am not sure who you have discussed DIDSON (Dual Frequency Identification Sonar) with, but I would not consider it voodoo science. I am actually kind of interested who you were speaking with regarding DIDSON and their level of experience with the unit as well as post-season analysis of the data. I have worked quite intensely with DIDSON since 2006 on various sized rivers doing sockeye enumeration (i.e. Chilko, Horsefly, Mitchell and the Quesnel). I have been involved site selection, set-up, inseason operation, demobilization of the project and post-season data analysis. I am also in constant consultation with various users of the unit as well as the manufacturer. In a nutshell, I am a very reliable source of information regarding DIDSON. DIDSON was technology once used by the US Navy to look for suspicious things planted in harbours. The technology was later seen as having other beneficial uses – such as fisheries work, so when the military released the technology various applications soon followed. DIDSON is used by various organizations now including DFO and the Alaska Dept. of Fish and Wildlife, so it is not a specific DFO voodoo machine (lol). Even the Americans use it for their enumeration projects.
Second, you mention double counting. As you can understand migrating salmon generally head upstream to spawn, but can go downstream for various reasons. Downstream movement at a DIDSON site is not handled that much differently from fish going back downstream through a fence. As with a fence, you are interested in the net flux (i.e. 5 go up….3 go back down…so the net is 2 upstream) past a point. With DIDSON software we are able to playback the recorded files like a VCR or DVD player. I can slow down the frame rate when things get busy so I can key in on individual fish and count them once. Double counting could occur if the observer chooses too high of a frame rate to count a large group of sockeye, but our crews are trained not to do this. We have protocols and procedures in place for counting DIDSON files properly.
Normally, downstream movement does not present that much difficulty (see site selection below) to our projects; however, when you have excessive downstream movement (i.e. hundreds in less than an hour lets say) it can present a problem, BUT this is much more of a problem if these fish mill in front of the sonar. This can happen when you choose the wrong site – like directly on the spawning grounds. It is like looking into your fish aquarium and trying to keep track what is in, what is staying and what is leaving. This is why we spend a great deal of effort trying to find the best site possible for DIDSON.
When we look for an optimal DIDSON site we attempt to satisfy behavioural, acoustic and human factors. It is not always possible to satisfy them all equally because Mother Nature and logistics play a role, but for the most part we do a very good job. We attempt to find sites that have the following characteristics: below all known spawning (very important…probably the biggest); a section of river where fish are going to want to migrate (i.e. It has to have some flow….A pool is no good because sockeye will hold in pools prior to spawning and mill around); the substrate cannot have too many boulders that cause blind zones or create turbulence that can create “noise” and effect counting conditions; a section of river where a weir can be built (we don’t block off the whole river….we have a fairly large opening where fish can pass through unlike a traditional counting fence); and an area that does not have much downstream movement (some areas have more downstream movement than others such as spawning areas). Prior to any installation, numerous candidate sites are visited and examined using the criteria including past enumeration knowledge (i.e. where fish have spawned previously and any history of excessive downstream movement).
As you are aware, many of the rivers we enumerate have human activity (i.e. angling), so we have to take that into consideration – not just because they can affect fish behaviour near the DIDSON, but also for safety reasons. For instance, some of the rivers where we use DIDSON have guided fishing which can utilize jet boats and rafts. Any site chosen needs to take this into consideration because we do not want to locate a weir at a blind corner in the river. The last thing we want is some boat crashing into it, so the site ideally has to allow boaters enough time to identify the weir and make the necessary corrections in direction (signage is installed upstream and downstream of ALL our fence and weir installations to aid the public).
It is true that discerning between species of fish with DIDSON is not an exact science; however, that certainly does not mean that DIDSON cannot be used successfully for enumeration. At one of our sites (Chilko) that we use DIDSON, we have observed a spatial segregation between Sockeye and co-migrating Chinook. Chinook tend to stick to the centre of the channel while Sockeye migrate close to the banks – many are within 2 feet of the shore. There are some morphological and energy reasons for this if you look into it further (I don’t have the reports handy with me at the moment, but they do exist). I spend quite a bit of time watching fish behaviour from the weir while doing DIDSON. The vast majority of the fish migrating at the time are sockeye and the occurrences of smaller (jack) Chinook in this river are very low to almost nil, so it is actually not a problem many may believe. There are also significant size differences between Sockeye and Chinook; and Chinook do not tend to migrate in larger groups like Sockeye. Other species (i.e. Rainbow Trout and Bull Trout) are generally smaller than the Sockeye in this river. Lastly, approximately 95% of the Chinook spawning is downstream of the DIDSON site (3 km downstream of the site). The majority of the Sockeye spawning is immediately downstream Chilko Lake. Add all these together and the influence of co-migrating Chinook is negligible. If you were to come out to this site with me you would see what I mean. I even have recorded footage which demonstrates the segregation – it’s like 2 separate highways.
Despite this, we still do correct for species identification by doing live visual counts from the weir in case we need to post-season. Other systems (not the example above) with multiple species can present problems – I won’t deny that; however, you have to look at the goals of the project, the relative numbers of what you are looking at and what is actively migrating at the time. We also do a live count and carcass recovery upstream of the site. If we were several thousands of fish over or under on the DIDSON we would see it from the ground or air– especially during peak of spawn as these fish will be on redds .
Third, you mention observer error in your last reply. Observer error is not unique to DIDSON. In fact, any enumeration method (not just with fish) has some degree of observer error. The challenge is how you deal with. With DIDSON, we double count a number of files daily to get precision bounds to our estimate. Over the years we have being doing DIDSON I can proudly say that our precision between observers is extremely good. The key is to get everyone involved with the project on the same page (i.e. not counting sockeye at ridiculously high frame rates). This involves training crew members and having good crew supervisors and biologists who have the experience to check the data and crew performance inseason. DIDSON counts have been comparable to fence counts (ours and others) which are the most absolute count you can achieve. A fence is only as good as the crew operating it and the stability built into it to make sure it does allow fish to escape through without being accounted for; however, we conduct very reliable fence operations in my opinion. I have worked at many of our fence locations.
Fourth, you mentioned limits to counting across an entire channel. You need to remember that DIDSON is a very good, high precision enumeration tool, but it is not for every system we enumerate on. In addition to the other criteria mentioned above, in order to effectively use DIDSON we need to weir off part of a river to allow for a small window of passage (as small as 10 m in some locations). The weir basically funnels the sockeye through the ensonified area (where the sonar is). We ensure that the sonar covers this opening. If we cannot construct an effective weir on a river then DIDSON is not going to be a good method. The exception is Chilko. At Chilko the weir does not have to span the entire river because the vast majority of the sockeye migrate very close to the banks. The weir essentially pushes the sockeye out so that they are not directly in front of the transducer of the DIDSON. It is difficult counting when something is directly in front of your face. At Chilko, two DIDSONs are used– one on the right bank and one on the left.
Lastly, you mention about data capacity constraints limited to only counting 20 minutes out of one hour. I realize that many would see this initially and be amazed how this can work effectively to enumerate sockeye; however, there are proven, statistical reasons for sub-sampling 20 minutes. Research has shown that as you increase the sub-sampling over 20 minutes you do not significantly increase the accuracy and precision of the estimate. Basically, you get the best bang for your buck with 20 minutes.
DIDSON files utilize a great deal of hard drive space. Just an hour DIDSON file is over a gigabyte. This can put a strain on storage capacity in the field, but it is possible to record the entire hour file and count every fish in it. However, one has to question is it really necessary to increase the sub-sampling time or count the entire hour when statically it has been demonstrated not to be that much more of a benefit. The 20 minute count is expanded for the entire hour by multiplying it by 3. It may seem strange from the outside, but in the end it all evens out. Some files will overestimate while others will underestimate. Precision bounds (95% confidence limits) are already put on the estimate. Is it really important to say that there are 520,000 fish compared to 521,000 or 519,500 fish? At projects where we need to account for every Sockeye (i.e. Cultus) we operate intensive fence operations at established sites where every sockeye is counted.
I do concede that for some species, such as Chinook, you would likely need to consider sampling more than 20 minutes (even an hour) because the passage of these fish past the site can be quite low (you would want to see every occurrence) compared to Sockeye , so there is flexibility built into a DIDSON project should the need arise. Everyday during the migration period we record an hour long file (chosen at random) and count every Sockeye in it so we can compare it to the 20 minute sub-sampled file from the same time period. Although we are confident with our sub-sampling we do this to ensure data quality. It also shows due diligence.
One thing is that DIDSON also does is minimize our impact (water and land) on the systems we use it.
Thanks for creating this opportunity for open dialogue. See you on the flow.
really appreciate the time, energy and passion in the comments. Great to get an in-depth look from the inside.
i still have lots of questions about effectiveness… however, more concerned about when the last few straggling salmon swim past some of these capital intensive counting systems.
would be much happier to hear of money going into habitat rehabilitation… but so be it.
thanks again.